Elections, Policy Preferences and International Financial Market Constraints

Thomas Sattler University College Dublin thomas.sattler@ucd.ie November 4, 2010

Abstract Despite strong pressures from international nancial markets, political parties continue to propose policies that dier considerably. This article examines to what extent actual government policies have converged notwithstanding the distinct policies that parties embrace in their party manifestos, and whether potential policy convergence can be traced back to nancial openness. To answer these questions, the study analyzes stock market reactions to elections in industrialized democracies since the 1950s. If actual policies of government parties continue to diverge, stock prices should increase after a success of a nominally market-friendly party. Moreover, stocks should respond more strongly the more open the economy is to nancial ows because capital has better exit options. If actual policies have converged, the impact of nominal policy positions on stocks vanishes in more open economies because actual policies are similar for all governments. The results imply that overall policies have weakly converged.

Paper to be presented at the Conference of the International Political Economy Society (IPES), Harvard University, Boston, November 12, 2010. 1

Introduction

Despite the strong pressures from international nancial markets on governments abilities to choose freely among dierent economic and other policy options, political parties, at least nominally, continue to propose very distinct policy solutions. Parties, when running for oce, claim that they will implement policies that dier signicantly from those of their competitors. The proposition of distinct nominal policy positions still is an important component of democratic electoral competition, although observers often object that government parties pursue very similar actual policies once they assume oce. This study examines to what extent policies that parties implement once in power dier as much as they claim in their party manifestos. It also examines whether potential policy convergence can be traced back to pressures from international nancial markets. To answer these questions, I assess the reaction of domestic stock markets to election outcomes in industrialized countries since the 1950s, and how this reaction has changed when economies have become more open to international nancial ows. Stock market behavior reects the aggregate expectations of well-informed actors about actual (as opposed to nominal) policies. It is therefore possible to infer to what extent expected policies of governments are consistent with their nominal dierences. Stock markets should respond more strongly to elections if governments actual policies continue to diverge in nancially open economies. In contrast, stock markets reactions should decrease if governments actual policies have converged. By analyzing stock market responses to elections, it is possible to draw conclusions about convergence in a broad range of policy elds. Stock reactions reect expected decisions by the incoming government in very dierent policy areas that have an eect on the protability of rms. Protability of rms does not only depend on economic policies, but also on policies in other elds where a race to the bottom or convergence was predicted by critics of globalization. These policies include economic policies like scal policy (Basinger and Hallerberg, 2004; Swank and Steinmo, 2002) or monetary and exchange rate policies (Bearce, 2007; Broz and Frieden, 2001; Frieden, 1991), but also dierent types of regulatory policies like environmental regulation and consumer protection (Bernauer, 2003; Vogel, 1998). The 2

analysis therefore presents a broad picture that covers many of the most important policy tools that the elected government has at its disposal. The results suggest that overall policies have weakly converged. Greater nancial openness does not lead to stronger stock market responses to elections although the reallocation of capital is less costly in more open economies, which implies that policies by dierent governments have become more similar. At the same time, the stock responses do not disappear in nancially open economies, which suggests that actual policies by governments cannot have fully converged. We can conclude from these results that elections and democratic competition have become somewhat less important for returns on nancial investments in industrialized countries compared to the past. Nonetheless, elections have not become unimportant for investment decisions because it has become less costly to adjust portfolio investments after an election as nancial openness has increased. The results also complement recent research on politics and nancial markets in a number of ways (e.g., Bernhard and Leblang, 2006). First, it adds a comparative perspective to the growing research on nancial market responses to elections that, with very few exceptions, examines the nancial eects of few and carefully selected cases (e.g., Bechtel, 2009; Herron, 2000; Leblang and Mukherjee, 2005). The basic results are consistent with previous research on the political economy of stock markets showing that stock prices increase when a more right-wing government is elected and that they drop when a more left-wing government wins the election. Unlike previous research, this study unveils a strong mediating eect of political constraints. While the eect of ideology on stock market behavior is large in countries with a low degree of political constraints, this eect fully disappears in countries with a high level of constraints. This is because policy volatility is small in high-constraint countries (Henisz, 2004), and nancial investors anticipate this when interpreting election results. Finally, the analysis also shows that stock markets continue to respond to political processes for several weeks after an election until the government formation process is terminated.

2
2.1

Stocks and Politics

Background

In the most general terms, the behavior of stock markets represents the aggregate expectations of market participants about the future performance of rms. Since a stock price can be dened as the sum of discounted expected future dividends, an increase in a rms stock price reects the expectation that future dividends will raise, which in turn depends on the rms future protability. A negative return (i.e. a drop in the price for a stock) thus means that market participants infer from newly available information that the future protability of a rm will decrease. Positive returns (i.e. increases in the price of a stock) reect the expectation that the protability of a rm will increase. Besides the rm-specic characteristics inuencing its competitiveness, e.g. technological innovations, the expected protability of a rm to a considerable extent depends on government policies and political developments more generally. Governments have a wide variety of policy instruments available that they can use to aect economic and social outcomes. For instance, they can reduce taris, subsidize rms, increase the consumption of specic products or decrease the tax burden of rms. Alternatively, a government can impose a ban on military exports, increase corporate taxes or tighten regulations. All these policies will aect the protability of rms and hence should be reected in a rms stock price. Market participants have a private incentive to collect all available information about future government policies to be able to adequately judge the expected protability of a rm. Stock market reactions then reect the aggregate assessment of a large number of well-informed market participants about the expected actual policies by the future government. Accordingly, the recent political economy literature nds that stock prices tend to respond positively when the probability of a success of a market-friendly political party in an upcoming election increases (Bechtel, 2009; Fuess and Bechtel, 2008; Campello, 2007; Herron, 2000; Mukherjee and Leblang, 2007; Leblang and Mukherjee, 2004,

2005; Santa-Clara and Valkanov, 2003; Snowberg, Wolfers and Zitzewitz, 2007a,b).1 Since stocks respond strongly to policy decisions, their behavior after elections also allows us to make inferences if and how policymaking has changed with greater nancial openness. A common fear is that nancial pressure will force governments to engage in a race to the bottom or at least to converge to a common policy in a particular policy eld (Tanzi, 1996). Others contend that political parties have continued to implement distinct policies despite international pressures on governments and that these political dierences may even be a precondition for openness (Boix, 1998; Garrett, 1998). The two dierent views have dierent implications for the impact of openness on policymaking and therefore on stock reactions as will be discussed in detail below. The examination of stock markets thus is one way to examine to what extent actual policies of governments have converged, and, if convergence exists, whether it can be traced back to nancial openness. An advantage in this context is that stock prices represent expectations about a broad range of policies that have played an important role in the globalization debate. While other nancial markets, such as bond or foreign exchange markets, mostly reect inationary expectations following from the broad economic policies, stocks also incorporate information from other, more specic policy areas. They not only respond to the general economic policies, like monetary and exchange rate policies, that have been examined by previous research on policy convergence (Bearce, 2007; Frieden, 1991; Mosley, 2000), but also react to other policies for which a race to the bottom has been predicted, e.g. tax policy (Basinger and Hallerberg, 2004; Swank and Steinmo, 2002; Swank, 2006) or regulatory policy (Vogel, 1998; Bernauer, 2003; Bernauer and Cadu, 2004). These policies play an essential role in electoral competition and seriously aect the protability of rms. With few exceptions, notably Bernhard and Leblang (2006, esp. chapters 3 and 4) and Campello (2007), the studies on stocks and politics are restricted to one or few election(s) in single or two countries. To my knowledge, none of them explicitly compares open-economy to closed-economy periods.
1

2.2

The Eect of Financial Openness

The eect of nancial openness on policies and thus on stocks depends on two aspects. First, it depends on the governments economic cost of implementing less market-friendly policies, which, in turn, is aected by the costs of nancial investors to reallocate their investments. Second, it depends on parties political costs of converging towards the same policy position. The implications dier depending on the relative impact of openness on the two types of costs. In this section, I rst discuss how nancial openness aects the economic costs of less market-friendly policies and then elaborate how the implications change as we assign greater importance to the political costs of policy convergence that government parties may face. Before the mid-1970s, cross-border capital ows were costly because capital controls were an essential part of the Bretton Woods System (Eichengreen, 1996, chapter 4). Capital controls can take dierent forms, but all of them increase the costs of reallocating capital internationally.2 In the extreme case when international nancial exchange is forbidden, capital can only be transferred illegally, which would be prohibitively costly.3 For the other end of the spectrum when no restrictions exist, the costs of reallocating capital are reduced to the fees for international nancial services. The extent of nancial regulations mediates the benets of reallocating capital in response to expected unfavorable policies after an election. After the breakdown of the Bretton Woods System, capital controls were continuously reduced, and most industrialized countries were fully open to international capital ows by the early 1980s. With greater openness, the cost of reallocating capital decreased signicantly facilitating transfers of funds abroad when an election outcome is unfavorable for rms expected protability. Although this de jure openness is crucial for international capital ows, the costs also depend on other, technical factors, such as the existence of a welldeveloped international nancial infrastructure. The quick and easy reallocation of portfolio Capital controls can include quantitative restrictions on international nancial exchange, administrative rules, e.g. the requirement to seek the approval by authorities, or taxes on international transactions. 3 This extreme case does not apply to the countries analyzed in this study, but mostly is the case for countries where free political competition and elections and well-developed stock markets do not exist. 6
2

investments requires the internationalization of banking, e.g. by establishing branches in dierent countries, which provides channels to transfer money across borders. De facto nancial openness thus may follow de jure openness with a delay.4 When the costs of international nancial transactions decrease, the economic costs of governments of a market-unfriendly strategy increase. If governments implement policies that have adverse eects on the protability of rms, these rms will sooner or later relocate to an economy with a more market-friendly political environment. Increasing cross-border portfolio investment ows speed up and intensify the negative consequences of economic policies on long-term investments and domestic economic activities. If investors expect that the new economic policies decrease the protability of rms, they will reshue their portfolio and reduce investments in stocks that are adversely aected by the policy change. If the new governments policies negatively aect large parts of the economy, this will lead to a considerable net outow of short-term capital reducing the overall capital stock in the country. Ultimately, a reduction in the capital stock of a country means that investments and therefore economic growth slow down. A government thus may be very cautious to adopt potentially adverse economic policies in an open economy. Instead, most political parties, once in power, may nd it preferable to converge on similar actual policies, for two reasons. First, voters may punish the government for poor economic performance when the negative eects of government policies on the economy materialize. Second, capital outows have direct repercussions on the governments position because public revenues will decrease restricting the governments ability to implement its preferred policies in other domains. Despite these nancial market constraints and convergence of actual policies, it may still be reasonable for political parties to adopt nominally distinct economic policy positions. Empirical data show that a signicant increase in actual cross-border portfolio nancial ows occurred only about a decade or more after de jure restrictions were abolished. The Quinn-Toyoda measure of de jure capital openness (Quinn, 1997; Quinn and Toyoda, 2008) indicates that most industrialized countries were almost fully open by 1980, but portfolio investment ows increased only slowly during the 1980s and exploded during the 1990s (Lane and Milesi-Ferretti, 2006). 7
4

Given the complex economic relationships in open economies, the parties in power could benet politically, at least in the short term, by claiming that favorable economic outcomes are related to their distinct policy program (Sattler, Freeman and Brandt, 2008; Sattler, Brandt and Freeman, 2010). Similarly, an opposition party may benet from blaming the government parties and their programs for poor economic performance and oer a nominally dierent policy proposal. This argumentation implies that, when the economic costs of non-convergence increase with greater nancial openness, governments will pursue increasingly similar policies that directly and indirectly aect the protability of major rms or large segments of the economy. If actual government policies have converged when countries became more open to international nancial ows, domestic stocks should respond less to elections when international nancial integration increases. Similarly, the eect of nominal ideology of the winning parties on stock markets should decrease as nancial openness increases. This strong convergence logic is summarized in the following two hypotheses.

H1a: As nancial integration increases, stock market responses to elections decrease. H1b: The negative (positive) response of stock markets to the election of a left-wing (right-wing) country becomes smaller as nancial integration increases.

Besides these economic considerations, the government also has to take into account the political costs that arise when it simply follows a market-friendly strategy and neglects the political demands by voters. Workers usually strongly resist cuts in social programs that insure them against poverty in case of unemployment (Rodrik, 1998). Organized labor also tends to oppose intentions to loosen labor regulations which would allow rms to oset employees more easily, at least those that are represented by powerful interest groups (Rueda, 2007). Similarly, employees in import-competing sectors tend to demand protection or compensation when competition from abroad increases (Hiscox, 2002; Rickard, 2007). All these policies aect rms protability because they require higher taxes on prots, limit the ability

to reduce excess production capacities or raise input prices for other sectors. Political pressure from voters and interest groups thus constrains the ability of governments to fully comply with the interests of nancial investors. Parties in power would lose important votes if they simply followed the interests of market participants and were unresponsive to demands from other groups. Since parties cannot aord to systematically ignore their political constituencies, they have to evaluate the economic costs of less market-friendly policies against the political costs of neglecting the interests of the societal groups that support them. This means that the greater economic costs of distinct policies in open economies may lead to more pro-market policies, but the policies of parties will not fully converge. In such a weak convergence scenario, governments balance their political and economic costs, which leads minor or no changes in stock market responses to elections as nancial openness increases. This is the case because, on average, expected government policies become more market-friendly, which leads to less reallocation of nancial funds when a new government comes into oce. At the same time, this reallocation becomes less costly, and it is optimal for market participants to reshue their portfolios even though expected policies change less than in closed economies. We thus should see only little eects of nancial openness on the relationship between stocks and politics according to this weak convergence logic.

H2a: As nancial integration increases, stock market responses to elections do not change. H2b: The negative (positive) response of stock markets to the election of a left-wing (right-wing) party remains constant as nancial integration increases.

Finally, some research implies that the political costs of converging towards the same policies increase with greater nancial openness. When the social and economic risks of voters increase as economies become more open, the political pressure to compensate them for these risks becomes larger (Garrett, 1998). The guarantee that (some) political parties will ensure the continuation of key social programs is even considered a political precondition for economic openness. This is because voters would not agree with the reduction of economic

barriers if they were not suciently insured against the greater risks that are associated with these policies (Adser` and Boix, 2002). a The social consequences of globalization may also lead to a more heterogenous electorate potentially increasing the costs of political convergence for parties representing dierent constituencies. New social cleavages can aect political competition in dierent ways, but their eect on stock markets is similar. Mainstream parties can take more diverging positions to represent the interest of the dierent groups, or small niche parties will exploit the electoral potential of winners and losers of globalization (Kriesi et al., 2008). In any case, governments that rely heavily on the support from those who are aected most by globalization will not be able to ignore demands for greater social insurance and have to accommodate these pressures when once in oce. In contrast, governments dominated by parties representing people who benet from economic integration will prefer to resort to more market-friendly policies. When both the economic and the political costs increase with greater nancial openness, policies remain largely the same. Since investors have better exit options when the domestic nancial market is well integrated into world nancial markets, they can adjust their portfolios more easily in response to an unfavorable election outcome. If a pro-market party wins the election, foreign investors may consider investing in the domestic stock market because expected protability of rms and thus expected returns on capital increase relative to foreign markets. This third, non-convergence logic implies that, stock markets should respond more to elections in nancially open economies than in nancially closed ones.

H3a: As nancial integration increases, stock market responses to elections increase. H3b: The negative (positive) response of stock markets to the election of a left-wing (right-wing) country becomes large as nancial integration increases.

10

3
3.1

Empirical Design
Event Study

To examine whether and how stock markets react to election outcomes from a comparative perspective and over time, I use an event study methodology (Binder 1998; Campbell, Lo and MacKinley 1997, ch. 4). The event study compares the behavior of nancial markets after an event to the expected behavior if that event had not happened. In other words, it assesses the dierences in stock market behavior after an election relative to normal stock behavior in non-election periods. This dierence between expected and actual behavior is the abnormal return. The abnormal returns then can be cumulated over a pre-specied period to provide an estimate of the impact of elections on nancial market behavior. If the resulting cumulative abnormal returns (CAR) are large and statistically signicant, we can conclude that the election has a considerable impact on nancial investment decisions. For instance, if stock markets drop more than expected in response to an election, i.e. the CAR is negative, this suggests that market participants expect a shift in government policies that is unfavorable for rms in the market. If stocks increase more than expected, i.e. the CAR is positive, market participants expect that the policies by the new government will have positive eects on rms protability. Figure 1 shows the logic of an events study graphically. To compute expected returns, I use a statistical model that estimates the relationship between the returns of a domestic stock index and a leading foreign market index:

ri,t = ai + bi rm,t1 + ei,t

(1)

The model is estimated for the estimation window, T0 t T1 . To avoid that the electoral campaign confounds the results from the estimation window, the window ends 50 trading days before the election day, i.e. T1 = 50 assuming that the election takes place at date 0. The length of the estimation period is exactly 100 days, i.e. T0 = 150. The market

11

index should be one that predicts movements in the domestic market well. I use the Dow Jones as market index for all countries. Other market indices were also considered, e.g. the DAX or the FTSE, but explanatory power of domestic nancial market behavior was not systematically greater using these other indices. The market index is lagged by one day because U.S. stock markets close later than markets in the countries analyzed. Empirical tests show that the specication with the lag is preferable to the one without lag.5 Figure 1: Time Line Estimation Window Event Window

T0

T1

T2

To compute abnormal returns during the event window, which is the period surrounding an event, we can calculate the deviation of actual returns from the predicted returns using the estimates of (1). Specically, suppose ei,t is the abnormal return for period t with T1 < t T2 , then ei,t = ri,t ai i rm,t1 b (2)

Abnormal returns can be accumulated over a pre-specied period from 1 to 2 within the event window, T1 < 1 2 T2 , to examine the impact of an election on stocks:
2

CARi =
t=1

ei,t

(3)

Since I am interested in the response of stock markets to the outcome of an election, abnormal The simpler constant-mean-returns model that sets bi = 0 is also used. Consistent with Campbell, Lo and MacKinley (1997), the results are very similar. Alternatively, one could use a more complicated, economic model, but the gains of such more complex models are small in practice (Campbell, Lo and MacKinley, 1997, 156/157). 12
5

returns are cumulated for ve subsequent trading days (or, a trading week) starting on the election day, i.e. 0 = 0 and 1 = 4. The start on the election day (as opposed to one day after the election) is motivated by the possibility that some information about the election result is already available on the day of the election. For further empirical analyses, CARs for longer windows, up to 60 trading days after the election, are also computed. These extended event windows show how persistent the eect of elections on stocks is and how sensitive the results are to the denition of the event window (Snowberg, Wolfers and Zitzewitz, 2008) Figure 2: Distribution of CARs

This procedure is repeated for each election in industrialized countries for which daily stock market data are available. This yields the dependent variable, which is the estimated CAR, the measure of stock market reactions after each election. As Figure 2 shows, the distribution of the 5-day CARs from the market model is very close to a normal distribution, with the exception of a disproportionate number of CARs around zero. The probability of a CAR around zero is large for three reasons. First, some elections are well predictable in which case stock markets respond little after an election (Snowberg, Wolfers and Zitzewitz, 13

2008). As a solution, it is possible weight the CARs by the amount of surplus votes that a majority government wins. This procedure in fact reduces the weight of CARs around 0. But there is no equally obvious way to account for predictability of elections that yield a minority government, which means that this procedure only works for a subsample of the dataset. I therefore use the unweighted CARs for the main analyses keeping in mind that they yield conservative estimates of the relationship between elections and stocks. Second, it may not be fully clear which parties will form the new government immediately after an election in a parliamentary democracy (Bernhard and Leblang, 2006, esp. chapters 3 and 6). If the distribution of CARs is adjusted for government formation time, the weight of CARs around zero is further reduced. I will therefore assess the impact of government formation processes when analyzing the extended event windows (up to 60 trading days or roughly 3 months). As expected, the eect of elections on stocks increases for longer events windows that capture the government formation process. It should be noted that excessively long government formation processes are rare in my dataset. The average government formation time as measured by the number of days from the election to the inauguration day is only 34 with a standard deviation of 31. Median formation time is only 25 days.6 Third, stock market participants may not be interested in elections in political systems with high political constraints because policy will not change when a new government assumes oce. I will explicitly model the role of political constraints in my empirical analyses.7

3.2

Countries and Data

The analysis covers elections in all parliamentary and semi-presidential industrialized democracies for which the relevant data are available. The analysis starts in the 1950s, but daily stock index series start later for many countries. As a basic rule, the broadest available Moreoever, government formation time at the 90th percentile is 67 days. Considering that the government composition is known well before the inauguration day, this is not a particularly long period of time. 7 Government formation and political constraints may be related, but they are conceptually dierent aspects. I therefore treat the two separately.
6

14

stock index that goes as far back in time as possible is used. In some countries, stock indices that exist for early years are not continued until the end of the analysis. In this case, other indices that do not exist for early years must be used for the later years. Generally, there is a fairly large period when the dierent indices overlap. I do multiple tests to see whether the dierent indices represent the same stock market developments. First, I examine the daily returns of the dierent indices. Regressing returns of one index on returns of the other always yields a very strong statistical relationship between the dierent indices. Second, I compare the estimated abnormal returns and CARs for the overlapping periods. The CARs from the dierent indices generally are extremely similar, often almost exactly identical. Data on election dates are from the Comparative Manifesto Project (Budge et al., 2001; Klingemann et al., 2006). To compute abnormal returns after an election, a domestic stock index must be available for at least 150 trading days before the election. This is the case for 205 elections. Six elections in the dataset took place in the 1950s, and 71 out of the 205 elections are from the 1950s, 1960s or 1970s, all decades when domestic nancial markets were fairly closed compared to today. Government ideology is measured using the left-right index from the comparative manifesto project. The index contains detailed information about nominal policy positions, and also takes into account the weights of individual parties involved in the government. The index thus captures as much information about government characteristics as possible, which is important for an analysis of nancial markets. Two dierent indicators of nancial openness are used. The Quinn-Toyoda indicator and its extensions (Quinn, 1997, 541-547; Quinn and Toyoda, 2008, 1409-1411) captures the degree of de jure capital account openness, i.e. the degree to which the government formally restricts cross-border capital ows. The second indicator is the log of actual portfolio investment ows as a percentage of GDP (Lane and Milesi-Ferretti, 2006) because the pressure on governments not necessarily arises from de jure, but from de facto economic integration.8 These data only exist from 1970 onwards, but actual ows were very small in the 1970s, For robustness checks, I also use total nancial ows, which is the sum of aggregate nancial assets and liabilities and also includes foreign direct investments.
8

15

increased only slowly during the 1980s and exploded during the 1990s, which means that there is considerable variation in actual ows for an analysis starting only in 1970.9 A number of variables are used to measure the degree of political constraints that a government faces, specically minority and multi-party governments, central bank independence and a more comprehensive constraints index that captures a variety of checks in a political system. Data on minority and multi-party governments is from Woldendorp, Keman and Budge (2000). Data on central bank independence are from Cukierman, Webb and Neyapti (1992) and the update by Crowe and Meade (2008). For a more comprehensive measure of political constraints, I use the indicator by Henisz (2002). This index is designed to capture the ability of policymakers to implement their most preferred policy. It takes into account veto points, the structure of the legislature, preference heterogeneity within the legislature and the role of government branches to assess the feasibility of policy change. Table 1 presents the summary statistics. For the ideology measure, higher numbers indicate more left-wing governments and is recoded to vary between 0 and 100 for this dataset. For the openness and political constraints variables, higher values indicate greater openness or constraints, respectively. To simplify the interpretation of the interaction terms, the most important intervening variable are recoded such that they vary between 0 and 10. Table 1: Summary Statistics Variable CAR Ideology De Jure Openness Log(Flows) Constraints Index Minority Multi-Party CBI
9

Obs 205 205 205 176 205 205 205 205

Mean 0.212 46.40 7.48 2.18 5.53 .243 .731 0.411

Std. Dev. 4.50 16.42 2.26 1.30 1.72 .430 .444 0.189

Min -16.10 0 0 0 0 0 0 0.09

Max 18.42 100 10 6.00 10 1 1 0.85

The original, not log-transformed indicators yield the same results. When the results dier, they depend on a few outliers, generally small countries with exceptionally large levels of nancial ows, such as Ireland.

16

Empirical Specications and Expectations

The analysis how ideology and nancial openness aect stock market responses proceeds in two steps. In the rst step, I will examine the eect of nancial openness on the size of stock market responses (hypotheses 1a, 2a and 3a). In the second step, I will analyze how nancial openness interacts with the eect of ideology and how they jointly aect the direction of stock market behavior after elections (hypotheses 1b, 2b and 3b). For the rst set of hypotheses, only the size of the stock market response matters because the hypotheses state that with growing nancial integration the size of the (negative or positive) CAR varies. The dependent variable thus is the absolute value of the CAR after an election, |CARi |. Besides the openness indicators, the specications include the control variables capturing how much the government is constrained to implement its most preferred policy. Suppose there are k constraining variables, then the specication takes the form
k

|CARi | =0 + 1 Opennessi +
j=1

j+1 Constrainti,j +

(4)

The variables are coded such that the coecients on the constraints variables, 2 through k+1 , should be negative. The coecient on the nancial openness variable, 1 , varies depending on the theoretical argument. The strong convergence logic suggests that this coecient is negative, while the non-convergence logic suggests that it is positive. According to the weak convergence logic, it should be zero. As the absolute CAR is bounded at zero, an application of the standard linear model has some drawbacks, especially because it may predict negative values, which are theoretically impossible. A truncated regression model is not appropriate in this context as it is built on the assumption that the observations are drawn from a distribution with values below zero, which just cannot be observed empirically. This is not the case here because the size of stock market reactions, as represented by the absolute CAR, cannot take negative values. To solve this issue, I also use an exponential model which accounts for the fact that the dependent

17

variable cannot be negative |CARi | =exp(0 + + i ) (5)

In practice, the linear model yields reasonable results for this analysis. The estimation results do not predict negative CARs for the observable range of the independent variables. The substantive results from the linear and the exponential models are the same. The second set of hypotheses suggests that both the size and the direction of the CARs depend on the incoming governments ideology and nancial openness, while the eect of ideology is mediated by nancial openness. For instance, according to the convergence logic, the market response to a left-wing government should be less negative in an open economy than in a closed economy, which can be modeled using an interaction term between the two variables. However, any relevant control variable representing major political constraints on the governments ability to adjust policy must also be interacted with ideology, which substantially complicates the model. This is because greater constraints lead to less positive CARs for a right-wing government and less negative CARs for a left-wing government. Suppose again that we have k constraining variables, then the second specication is CARi = 0 + 1 Ideologyi + 2 Opennessi + 3 Ideologyi Opennessi
k

+
j=1

(j+3 Constrainti,j + j+k+3 Ideologyi Constrainti,j ) (6)

+ vi

Including all potentially relevant constraints variables would yield a fairly complex model. It is unlikely that there will be sucient observations for each combination of values to draw meaningful inferences from the estimation results with this many interaction terms. To keep the model as simple as possible, those constraints variable(s) will be used which turn out to be the most relevant in the rst part of the analysis using equation (4). Naturally, the variable(s) that best explain the size of the stock market response should be the one(s) having the strongest intervening impact on the relationship between ideology and stock markets. 18

In equation (6), the constant, 0 , should be positive because it shows how stock markets respond to the election of a right-wing government (Ideologyi = 0) with no nancial openness and no political constraints (Opennessi = 0; Constrainti,j = 0 for j = 1, ..., k). The coecient on ideology, 1 , should be negative because stocks drop more / increase less the more left-wing the government is. The coecients on the constraints variables, j+3 for j = 1, ..., k, indicate how growing constraints diminish the positive market response to a conservative government and should be negative. Finally, the coecients on the interactions between ideology and constraints, j+k+3 for j = 1, ..., k, show how the impact of a more left-wing government changes as constraints increase. Since the negative eect of left-wing governments decreases as constraints increase, these coecients should be positive. Table 2: Expected Values of Coecients Across Empirical Models Expected Coecient Baseline 0 > 0; j+1 < 0 1 > 0 Model 1 Non-Convergence Weak Convergence 1 = 0 Strong Convergence 1 < 0 Baseline 0 > 0; 1 < 0; j+3 < 0; j+k+3 > 0 2 > 0; 3 < 0 Model 2 Non-Convergence Weak Convergence 2 = 0; 3 = 0 Strong Convergence 2 < 0; 3 > 0 Subscript j refers to constraints variable j in the respective model. k is the maximum number of constraints included in the specication. The theoretical discussion has diverging implications for the impact of openness. The non-convergence logic suggests that openness should magnify the eect of ideology on stock market responses. This implies that stocks respond more positively to right-wing governments, but at the same time react more negatively to left-wing governments as openness increases, i.e. 2 > 0 and 3 < 0. The opposite is the case for the strong convergence logic saying that the impact of ideology diminishes as openness increases, i.e. 2 < 0 and 3 > 0. The weak convergence logic suggests that no intervening eect of openness exists, i.e. 2 = 0 and 3 = 0. Table 2 summarizes the implications of the theoretical discussion for 19

the empirical specications in (4) and (6).

5
5.1

Results
The Size of Stock Reactions

This rst section examines how nancial openness and political constraints aect the magnitude of the response of stock markets to elections. The results in Table 3 refer to equation (4). I examine the impact of the variables separately and then discuss the full models. The rst two columns show how increasing nancial openness aects the size of stock market reactions to elections. The results imply that the size of these reactions decreases as de jure nancial openness increases, but the coecient is not statistically signicant at conventional levels. This basic result holds for all other specications. The coecient is always negative, but not robustly statistically signicant. Replacing de jure openness with actual portfolio investment ows, we also see a negative eect of greater nancial ows on the size of stock market reactions, but again the coecient is not statistically signicant.10 The results are the same for specications controlling for political constraints as the last two columns show. The substantive eects of the de jure openness and actual ows indicators are very similar and small. On average, stocks change by 4.20% (3.88%) after an election for zero de jure openness (actual nancial ows). The coecient on the de jure (actual ows) indicator suggests that the stock market reactions drop by 1.47 (1.69) percentage points when openness increases from its minimum to its maximum. Overall, additional analyses yield the same conclusions. We get the same result when the exponential model is used or when portfolio ows are replaced with total investment ows. The results show a consistent negative eect, but this eect is rather small in substantive terms and not statistically signicant. These rst results favor the weak convergence hypothesis suggesting that government policies have converged somewhat, but far less strongly than predicted by the strong convergence thesis. The number of observations is smaller because this series is only available from 1970. See section 3.2
10

20

Table 3: Eects of Openness and Political Constraints on Stock Markets Flows Minority Multi CBI Index Full(1) -0.145 (0.153) Full(2)

Openness De Jure Openness -0.147 (0.141) Log(Flows) -0.282 (0.241) -0.289 (0.444)

Minority

21

-0.399 (0.504) Multi 0.343 0.904 (0.431) (0.454) CBI -0.131 0.021 (1.216) (1.477) Index -0.328 -0.377 (0.109) (0.117) Constant 4.196 3.878 3.169 2.847 3.152 4.911 5.695 (1.212) (0.752) (0.284) (0.316) (0.531) (0.761) (1.209) F 1.09 1.37 0.42 0.63 0.01 9.08 3.35 P rob > F 0.309 0.255 0.522 0.435 0.915 0.007 0.022 R2 0.01 0.01 0.00 0.00 0.00 0.03 0.05 N 205 176 205 205 205 205 205 Standard errors in parantheses. Errors cluster on countries

-0.139 (0.284) -0.776 (0.430) 1.217 (0.467) -0.631 (1.831) -0.479 (0.151) 5.831 (0.732) 4.90 0.004 0.07 176

The analysis in this section also explores which political constraints are most relevant from a nancial market perspective. The estimation results for the two government characteristics variables, minority and multi-party government, suggest that market participants do not rely on this information to draw conclusions about the size of expected policy change. None of the two coecients is statistically signicant. One interpretation is that these dummy variables are too crude as measures of political constraints. Financial market participants are likely to use more nuanced information than these two indicators provide. More surprising is the weak performance of central bank independence. The isolation of the bank from political pressure generally leads to lower inationary expectations and therefore to lower real returns on stocks. An explanation for this result is that stock markets (as opposed to bond markets) care more about expected scal, redistributive and industrial policies of governments that more directly aect rms protability than monetary policy which only indirectly aects the performance of rms. Finally, the estimated impact of the more general political constraints index conrms that market participants use more detailed information about the institutional and political checks that governments face. This index is the most comprehensive one and subsumes the simpler, previous measures of political constraints and includes additional aspects, e.g. the distribution of power in a second parliamentary chamber. The coecient on this indicator is statistically signicant for this and all other specications and shows that greater constraints lead to substantially lower absolute stock market movements after an election. Specically, in a low-constraints country, stocks on average change by 4.9%. For a high-constraints country, this eect decreases by 3.3 percentage points to 1.6%.

5.2

The Direction of Stock Reactions

This section examines in which direction stocks move after the election of a left- or right-wing government. The results in Table 4 are based on equation (6) or components thereof. The results in Table 3 suggest that for this section, only the general constraints index should be used. The other indicators and the resulting additional interaction terms would substan-

22

tially complicate the analysis without much analytical gain. Equation (6) then reduces to a specication with only one constraints variable, i.e. k = 1. The rst column reports the impact of the incoming governments ideology on stock markets. The estimate for the constant shows that if fully right-wing government wins the election, i.e. when ideology takes the value 0, then stock markets increase by 1.9%. As expected, the CAR decreases as the government becomes more left-wing, and this eect is statistically signicant. When we move from a fully right-wing to a fully left-wing government, the CAR drops by 3.7 percentage points. In other words, when a fully left-wing government comes into oce, then stock markets decrease by 1.8%. The almost exact symmetry between left- and right-wing eects, i.e. the result that stock prices drop by exactly the same amount after a left-wing success as they increase after a right-wing success, is astonishing, even more so as it is not expected. Pantzalis, Stangeland and Turtle (2000) show that stock markets on average increase after an election because the resolution of the electoral uncertainty itself leads to a positive reaction. Taking into account the content of the election outcome, we would expect that stocks rise more after a right-wing success than they drop after an equivalent left-wing success. As discussed above, the impact of ideology should vary with the political constraints that a government faces. The specication in the second column includes the political constraints variable and the interaction between constraints and ideology. The coecients on all variables and the constant show the expected signs. In a country with zero political constraints, the expected CAR accumulates to 5% when a fully right-wing government comes into oce. This is more than two and a half times as much as the expected CAR for the same ideology change when the intervening eect of constraints is ignored. When a fully leftwing government is elected in a low constraints country, stock markets decrease by 3%. The asymmetric eect of left- and right-wing success now also conrms the ndings by Pantzalis, Stangeland and Turtle (2000) discussed in the previous paragraph. To simplify the interpretation of the intervening eect of political constraints, Figure 3 shows the marginal eect of an increase in the ideology variable on CARs as political

23

Table 4: Eects of Ideology on Stock Markets

Ideology

De Jure -0.031 (0.031)

Flows -0.056 (0.027)

Constraints

Ideology*Constraints

Simple Constraints -0.037 -0.080 (0.016) (0.040) -0.595 (0.360) 0.009 (0.008)

De Jure Openness

Ideology*De Jure Openness

24

-0.204 (0.323) -0.001 (0.006)

Log(Flows)

Ideology*Log(Flows)

1.920 5.022 3.485 (0.861) (1.880) (1.654) F 5.40 3.97 9.47 P rob > F 0.030 0.022 0.000 R2 0.02 0.03 0.03 N 205 205 205 Standard errors in parantheses. Errors cluster on countries

Constant

Full(1) Full(2) -0.094 -0.134 (0.051) (0.052) -0.656 -0.907 (0.377) (0.547) 0.010 0.016 (0.008) (0.009) -0.270 (0.314) 0.001 (0.006) -0.690 -0.521 (0.694) (0.607) 0.003 0.001 (0.013) (0.012) 4.033 7.358 8.473 (1.613) (2.566) (3.215) 3.85 11.82 5.90 0.024 0.000 0.001 0.05 0.04 0.06 176 205 176

constraints increase. The graph thus shows how the market response changes when a more left-wing government is elected, for dierent degrees of political constraints. It conrms that stock markets drop in response to the election of a left-wing government. But when political constraints increase, the impact of government ideology decreases and becomes statistically insignicant at a value of ca. 5.5 on the political constraints scale. This level of constraints largely corresponds to Germany in the 1990s when the opposition parties dominated the second chamber of parliament, the Bundesrat, and the government was unable to implement major reforms without the consent of the opposition. When constraints increase further, the impact of ideology disappears. Such high levels of constraints more or less correspond to political systems like Belgium or Switzerland. Figure 3: Eect of a One-Unit Increase in Ideology as Political and Financial Constraints Increase

The eect of government ideology in low and medium constraints countries is sizeable in substantive terms. If, for instance, government ideology increases from one standard deviation below to one standard deviation above the mean, i.e. an increase by ca. 32

25

units on the ideology scale, the CAR decreases by more than 2.5%. For a comparison, when Mitterrand won against Giscard dEstaing in the 1981 French election, the government ideology variable increases by more than 45 units, which means that the predicted decrease in the CAR would be 3.6% in this example. Figure 4: Eect of a One-Unit Increase in Ideology as De Jure Openness Increases

The results from the nancial openness specications in columns three through six of Table 4 conrm the ndings from the previous section. Figure 4 plots the marginal eect of ideology as de jure openness increases. The marginal eect does not change much as de jure openness increases. It seems to decline slightly, which reects the negative coecient on the interaction between ideology and de jure openness in the third column in Table 4, but this eect is negligible. For the complex specication in the fth column that includes both interactions, the coecient on the ideology / de jure openness interaction is positive, which means that the marginal eect is increasing by the same amount as it is decreasing in Figure 4. The corresponding marginal eects for the actual portfolio ows are plotted in Figure 5. 26

Figure 5: Eect of a One-Unit Increase in Ideology as Financial Flows Increase

They suggest that stocks react considerably when a more left-wing party wins the election for low levels of nancial ows, but this eect becomes statistically insignicant for high levels of nancial ows. However, the widening of the condence intervals essentially occurs because estimation uncertainty is larger at the endpoints of the distribution. The decrease in marginal eects as represented by the solid line is substantively irrelevant. Again, I interpret these results as evidence in favor of a weak convergence hypothesis. To examine the intervening impact of nancial openness further, the sample was split in two parts: a pre-1980 and a post-1980 subsample. This procedure reects the idea that international nancial integration has increased massively since the 1980s, but was comparatively small before. This sample split represents a rough approximation to changes in nancial integration with the advantage that the model complexity can be reduced again. It allows me to estimate the simpler political constraints model for the two subperiods without using double interactions as in the combined model of Table 4. The results suggest that the eect of politics on nancial markets decreased from the pre- to the post-1980 period,

27

but again this eect is not statistically signicant. The marginal eects imply that the joint impact of ideology and constraints is larger for the pre-1980 period and smaller for the post1980 period than for the whole period. However, using a Chow test, I cannot reject the null hypothesis that the coecients are stable over time at conventional signicance levels. Moreover, a rolling regression using a sequence of shorter subsamples does not show that the eect is as strong and as systematic as the split in pre- and post-1980 periods suggests.

5.3

Additional Analyses

To examine how short-lived persistent the eect of elections on stocks is, and how sensitive the results are to the specication of the event window, I reestimate the models using CARs computed for event windows of dierent lengths, up to 60 trading days after an election. Figure 6 plots the marginal eects of ideology scale for dierent event windows. The eects are computed using estimation results based on the specication in the second column of Table 4. The political constraints variable is set at one standard deviation below the mean, which corresponds to a value of 3.81.11 The gure shows that the nancial eects of elections are highly persistent and that the results are not an artefact of a particular event window denition. The marginal eect is consistently negative and the 95% condence interval does not include zero for windows up to 40 days. The magnitude of the eect is even increasing and more than doubles during the rst 15 trading days (3 weeks) and remains roughly constant after that. This is plausible considering that it may take a few weeks until it is fully clear which parties will form the new government. Stocks continue to respond to political processes until the government formation process is nished (Bernhard and Leblang, 2006). Comparing the time frame during which stocks continue to respond to electoral politics with the average and median government formation time in the dataset conrms this. The average time between the election and the The eect for the 5-day window in Figure 6 corresponds to the marginal eect in Figure 3 where the constraints variable on the x -axis takes the value 3.81. This eect is -0.048. The constraints variable is set below the mean because the aim here is to show that the strong eect of ideology for low contraints levels is robust against changes in the event window and also persists for more than just a few days. 28
11

Figure 6: Marginal Eects for Dierent Event Window Lengths

inauguration day is 34 days (4.8 weeks), while stocks continue to react to elections for about 3 weeks after the election. The median formation time is only 25 days (3.6 weeks), which is even closer to the 3-week horizon during which stocks continue to react.12 It is also possible that stock market patterns unrelated to elections produce the results shown above. To address this question, I use a placebo analysis which repeats the empirical analysis for randomly chosen non-election days. If my ndings are in fact caused by elections and their outcomes, stock movements should not be systematically associated with political preferences and institutions when they were not treated with an election. To examine this conjecture, I draw 1000 random samples of non-election days of the same size as my election The duration of the government formation process is measured as the number of days from the election day until the inauguration day. However, the information which parties will form the government is available before the inauguration day implying that stocks should not respond during the full government formation period as it is dened here. The discrepancy between the mean and the median also indicates that in exceptional instances, the formation process took much longer, e.g. in Austria in 1999. These long government formation periods are unusual. Government formation duration at the 90th percentile is only 67 days.
12

29

dataset, which means that each sample covers 205 observations. Each sample is drawn from the set of days on which no election took place in the 22 countries that I analyze.13 Table 5: Comparison of Results from Main and Placebo Analyses Simple -0.037 Constraints -0.048 De Jure -0.035 Flows -0.053

Main Placebo

0.001 0.001 0.001 0.001 [-0.015; 0.017] [-0.024; 0.026] [-0.015; 0.018] [-0.017; 0.020] Table shows the marginal eects of ideology. Results in the rst row are from the main analysis in Table 4. Second row shows the median marginal eect computed from 1000 random samples of non-election days for the placebo analysis. The numbers in brackets are the 5 and 95 percentiles. For the placebo analysis, I reestimate equation (6) and compute the marginal eect of ideology for pre-specied levels of political constraints and nancial openness using each of the 1000 samples. The second row in Table 5 presents the estimated median marginal eect from the 1000 random samples for the dierent specications in Table 4. The numbers in brackets are the 5 and 95 percentiles. The marginal eect for the simple model in the rst column is equivalent to the coecient on ideology. The marginal eect for the constraints model is computed for political constraints set at one standard deviation below the mean. It therefore corresponds to the marginal eects presented for the analysis of dierent event window lengths in Figure 6. The nancial openness variables are set at the same values. For comparison, the rst row shows the equivalent eects from the main analysis in Table 4. The results clearly show that stock market behavior during non-election periods is fundamentally dierent from stock behavior after elections. The median marginal eects from the placebo regressions are almost exactly zero for all specications indicating that stocks are not expected to respond to political preferences during non-election periods. The variation The underlying set of days covers all days not within an election period and for which CARs can be computed. To compute the CARs, a stock index must be available for a suciently long period of time before and after the potential date. I exclude 100 days before and 50 days after an actual election to ensure that elections do not confound the results of the estimation or event windows. Using this procedure, the random samples can be drawn from 188527 possible days in the 22 countries. 30
13

in these eects is small with standard deviations between 0.011 and 0.015. Moreover, the eects at the 5 percentile do not even come close to those from the main analysis. They are not even or just half of the eects from the main analysis for all specications. Even the maximum marginal eects from the placebo analysis (not in the table) are always smaller in magnitude than those from the main analysis. The placebo analysis thus strongly suggests that stocks behave systematically dierent after elections than during non-election periods. Besides the variety of tests described in the text, including dierent specications, subsamples, empirical measures and event windows, the results are robust to other changes. The results do not change if CARs are computed using a constant-returns instead of a market model (see section 3.1). I also used dierent indicators of government ideology from the Comparative Manifesto Project, specically governments economic policy and welfare state preferences. As expected, the impact of the economic preferences indicator is stronger than the general left-right index used in the text, while the impact of the welfare state index is weaker, but these two more specic indicators are less reliable than the general ideology measure.14 I therefore rely on the general indicator in the main analysis. When interpreting these results, it should be considered that the empirical strategy under estimates the eect of elections on stock markets. Stock markets should mostly respond if they have not or not fully anticipated the election outcome. While only few elections are perfectly predictable, market participants have clearer expectations about the outcome of one elections than of another. Stocks will respond less strongly to the election if the outcome corresponds to market participants expectations because the impact of expected policy on rms protability is already included in prices on the election day. This implies that the implications that we drew from the results are based on conservative estimates rather than estimates that exaggerate the relationship between politics and nance. Finally, a popular objection against an analysis like this one is that nancial markets are not able to rationally anticipate future government policies or the results reect market The issue-specic Comparative Manifesto indices rely on only few categories, some of which are not covered by many party programs. These missings then possibly lead to inadequate measures of party positions on particular policy dimenions. 31
14

behavior somehow related to elections, but unrelated to future government policies. Absent a broadly accepted, rival model of stock market behavior that could explain the results of this study, it is dicult to assess this general critique. It is possible, however, to evaluate a few, more concrete alternative explanations. If market participants believe that other traders buy after a right-wing success and sell after a left-wing victory, for whatever reason, it would be reasonable to do the same and make a prot when stock markets increase or avoid a loss when stock markets decrease. A more behavioral explanation would be that happy people tend to buy things and stock market participants, being more conservative on average, buy stocks when when their preferred party wins the election.15 These and related explanations are inconsistent with the empirical results above. They can explain why stocks drop or increase after the election of a left-wing or right-wing government. But following these mechanisms, there is no reason why this eect should disappear when political constraints increase, as the results show. The strong and robust impact of political constraints on stock market behavior after elections indicates that the stock market response is in fact related to expectations about future policies because these constraints reduce the ability of a new government to alter existing policies. Stock market behavior should not be related to political constraints if it was not about expected government policies.

Conclusion

This study examines how nancial markets respond to electoral outcomes and how these responses have changed as industrialized countries have become more open to nancial ows. The results provide evidence in favor of a weak convergence thesis, which implies that governments continue to implement distinct economic and related policies, but these policies are more similar across governments than in the past. Elections therefore have become less important for returns on nancial investments in industrialized countries, although they have not become unimportant for investment decisions. Elections continue to play an important This mechanism in essence reects the idea that stock markets react to all kind of news that should be unrelated to stock markets, e.g. the victory of a national sports team.
15

32

role for investment decisions despite some policy convergence because investors can more easily reallocate their investments internationally when policies are expected to change. The analysis also extends research on the political economy of nancial markets in a number of ways. First, it adds a historical dimension that is necessary to fully understand how globalization has aected the role of politics and nancial markets since the Second World War. Second, it provides a comprehensive comparative analysis that is rare in this eld of research. The results show that the institutional structure of a country has a strong and mediating eect on the relationship between politics and stocks. Stock markets move strongly and persistently after an election, but this eect fully disappears in countries with high political constraints. Checks and balances limit political risk for investors as they seriously constrain governments discretion to alter policies. Therefore, nancial investors pay less attention to elections in countries with high constraints or low political risk. The constraining eect of institutions is major, especially comparing it to the modest constraints that follow from international nancial integration. This analysis approaches the globalization debate from a highly aggregate level because the scope of this study does not allow for distinctions between dierent sectors and rms. Future research should examine the implications of this debate for stock markets in greater detail. Analyses of stocks on the sectoral level can reveal to what extent import-competing sectors beneted from a particular election outcome, possibly at the expense of exportoriented rms. Political parties may also favor dierent strategies to cope with globalization, which have dierent implications for redistribution among economic sectors. Such distributional eects should be visible in stock responses to elections on the sector or rm level.

33

References
Adser`, Al` and Charles Boix. 2002. Trade, Democracy and the Size of the Public Sector: a cia The Political Underpinnigs of Openness. International Organization 56(2):229262. Basinger, Scott and Mark Hallerberg. 2004. Remodeling the Competition for Capital: How Domestic Politics Erases the Race to the Bottom. American Political Science Review 98(2):261276. Bearce, David. 2007. Monetary Divergence: Domestic Policy Autonomy in the Post-Bretton Woods Era. University of Michigan Press. Bechtel, Michael. 2009. The Political Sources of Systematic Investment Risk: Lessons from a Consensus Democracy. Journal of Politics 71(2):667677. Bernauer, Thomas. 2003. Genes, Trade and Regulation: The Seeds of Conict in Food Biotechnology. Princeton University Press. Bernauer, Thomas and Ladina Cadu. 2004. In Whose Interest? Pressure Group Politics, Economic Competition and Environmental Regulation. Journal of Public Policy 24(1):99 126. Bernhard, William T. and David Leblang. 2006. Democratic Processes and Financial Markets: Pricing Politics. Cambridge University Press. Binder, John J. 1998. The Event Study Methodology Since 1969. Review of Quantitative Finance and Accounting 11(2):111137. Boix, Carles. 1998. Political Parties, Growth, and Equality: Conservative and Social Democratic Strategies in the World Economy. Cambridge University Press. Broz, J. Lawrence and Jery A. Frieden. 2001. The Political Economy of International Monetary Relations. Annual Review of Political Science 4:317343.

34

Budge, Ian, Hans-Dieter Klingemann, Andrea Volkens, Judith Bara and Eric Tanenbaum. 2001. Mapping Policy Preferences: Estimates for Parties, Electors and Governments 1945-1998. Oxford University Press. Campbell, John Y., Andrew W. Lo and A. Craig MacKinley. 1997. The Econometrics of Financial Markets. Princeton University Press. Campello, Daniela. 2007. Do Markets Vote? A Systematic Analysis of Portfolio Investors Response to National Elections. Manuscript. Crowe, Christopher and Ellen E. Meade. 2008. Central Bank Independence and Transparency: Evolution and Eectiveness. IMF Working Paper 1999. Cukierman, Alex, Steven B. Webb and Bilin Neyapti. 1992. Measuring Central Bank Independence and its Eect on Policy Outcomes. The World Bank Economic Review 6(3):353398. Eichengreen, Barry. 1996. Globalizing Capital: A History of the International Monetary System. Princeton University Press. Frieden, Jery A. 1991. Invested Interests: The Politics of National Economic Policies in a World of Global Finance. International Organization 45(4):425451. Fuess, Roland and Michael Bechtel. 2008. Partisan Politics and Stock Market Performance: The Eect of Expected Government partisanship on Stock Returns in the 2002 German Federal Election. Public Choice 135(3-4):131150. Garrett, Georey. 1998. Partisan Politics in the Gobal Economy. Cambridge University Press. Henisz, Witold J. 2002. The Institutional Environment for Infrastructure Investment. Industrial and Corporate Change 11(2):355389.

35

Henisz, Witold J. 2004. Political Institutions and Policy Volatility. Economics and Politics 16(1):127. Herron, Michael C. 2000. Estimating the Economic Impact of Political Party Competition in the 1992 British Election. American Journal of Political Science 44(2):320331. Hiscox, Michael. 2002. International Trade and Political Conict: Commerce, Coalitions, and Mobility. Princeton University Press. Klingemann, Hans-Dieter, Andrea Volkens, Judith Bara, Ian Budge and Michael D. McDonald. 2006. Mapping Political Preferences II: Estimates for Parties, Electors and Governments in Central and Eastern Europe, European Union and OECD 1990-2003. Oxford University Press. Kriesi, Hanspeter, Edgar Grande, Romain Lachat, Martin Dolezal, Simon Bornschier and Timotheos Frey. 2008. West European Politics in the Age of Globalization. Cambridge University Press. Lane, Philip R. and Gian Maria Milesi-Ferretti. 2006. The External Wealth of Nations Mark II: Revised and Extended Estimates of Foreign Assets and Liabilities, 1970-2004. IMF Working Paper WP/06/69. Leblang, David and Bumba Mukherjee. 2004. Presidential Elections and the Stock Markets: Comparing Markov-Switching and Fractionally Integrated GARCH Models of Stock Volatility. Political Analysis 12(3):296322. Leblang, David and Bumba Mukherjee. 2005. Government Partisanship, Elections, and the Stock Market: Examining American and British Stock Returns, 1930-2000. American Journal of Political Science 49(4):780802. Mosley, Layna. 2000. Room to Move: International Financial Markets and National Welfare States. International Organization 54(4):737773.

36

Mukherjee, Bumba and David Leblang. 2007. Partisan Politics, Interest Rates and the Stock Market: Evidence from American and British Returns in the Twentieth Century. Economics and Politics 19(2):135167. Pantzalis, Christos, David A. Stangeland and Harry J. Turtle. 2000. Political Elections and the Resolution of Uncertainty: The International Evidence. Journal of Banking and Finance 24:15751604. Quinn, Dennis. 1997. The Correlates of Change in International Financial Regulation. American Journal of Political Science 91(3):531551. Quinn, Dennis P. and A. Maria Toyoda. 2008. Does Capital Account Liberalization Lead to Growth? Review of Financial Studies 21(3):14031449. Rickard, Stephanie J. 2007. Compensating the Losers: Evidence of Policy Responses to Globalziation from Congressional Votes. Paper Presented at the Meeting of the International Political Economy Society, Stanford University, November 9-10. Rodrik, Dani. 1998. Why Do More Open Economies Have Bigger Governments? Journal of Political Economy 106(5):9971032. Rueda, David. 2007. Social Democracy Inside Out: Partisanship and Labor Market Policy in Industrialized Democracy. Oxford University Press. Santa-Clara, Pedro and Rossen Valkanov. 2003. The Presidential Puzzle: Political Cycles and the Stock Market. Journal of Finance 58(5):18411872. Sattler, Thomas, John R. Freeman and Patrick T. Brandt. 2008. Political Accountability and the Room to Maneuver: A Search for a Causal Chain. Comparative Political Studies 41(9):12121238. Sattler, Thomas, Patrick T. Brandt and John R. Freeman. 2010. Democratic Accountability in Open Economies. Quarterly Journal of Political Science 5(1):7197.

37

Snowberg, Erik, Justin Wolfers and Eric Zitzewitz. 2007a. Partian Impacts on the Economy: Evidence from Prediction Markets and Close Elections. Quarterly Journal of Economics 122(2):807829. Snowberg, Erik, Justin Wolfers and Eric Zitzewitz. 2007b. Party Inuence in Congress and the Economy. Quarterly Journal of Political Science 2:277286. Snowberg, Erik, Justin Wolfers and Eric Zitzewitz. 2008. How Prediction Mar-

kets Can Save Event Studies. Manuscript, California Institute of Technology, http://www.hss.caltech.edu/ snowberg/papers/Snowberg0event Swank, Duane. 2006. Tax Policy in an Era of Internationalization: Explaining the Spread of Neoliberalism. International Organization 60(4):847882. Swank, Duane and Sven Steinmo. 2002. The New Political Economy of Taxation in Advanced Capitalist Democracies. American Journal of Political Science 46(3):642655. Tanzi, Victor. 1996. Globalization, Tax Competition, and the Future of Tax Systems. IMF Working Paper WP/96/141. Vogel, David. 1998. Trading Up: Consumer and Environmental Regulation in a Global Economy. Harvard University Press. Woldendorp, Jaap, Hans Keman and Ian Budge. 2000. Party Government in 48 Democracies, 1945-1998. Kluwer Academic Publishers.

38